Caffeine in Citrus flowers

Abstract

The allocation of purine alkaloids within citrus flowers was studied and found to be linked to anthesis, with 99% of the total flower caffeine confined to the androecium. The main alkaloid is caffeine accompanied by considerable (up to 30% of caffeine) concentrations of theophylline. In the anther, these purine alkaloids reach altogether a concentration of 0.9% dry wt which is close to the caffeine content of the Arabica coffee bean. The pollen alkaloid concentration is in the same range. Much lower but still marked concentrations were found in the nectar. A considerable breakdown of alkaloids during honey production is assumed. The biological significance of this particular secondary compound allocation as well as possible effects on the key pollinator, the honey-bee, are discussed.

Full text

Caeine in Citrus ¯owers
Josef A. Kretschmar, Thomas W. Baumann*
Institute of Plant Biology, University of Zu
Èrich, Zollikerstr. 107, 8008 Zu
Èrich, Switzerland
Received 11 August 1998; received in revised form 4 January 1999; accepted 21 January 1999
Abstract
The allocation of purine alkaloids within citrus ¯owers was studied and found to be linked to anthesis, with 99% of the total
¯ower caeine con®ned to the androecium. The main alkaloid is caeine accompanied by considerable (up to 30% of caeine)
concentrations of theophylline. In the anther, these purine alkaloids reach altogether a concentration of 0.9% dry wt which is
close to the caeine content of the Arabica coee bean. The pollen alkaloid concentration is in the same range. Much lower but
still marked concentrations were found in the nectar. A considerable breakdown of alkaloids during honey production is
assumed. The biological signi®cance of this particular secondary compound allocation as well as possible eects on the key
pollinator, the honey-bee, are discussed. #1999 Elsevier Science Ltd. All rights reserved.
Keywords: Citrus; Rutaceae; Caeine; Theophylline; Purine alkaloids; Flower development; Nectar; Pollen; Apis mellifera; Honey-bee
1. Introduction
The compilations often found in literature (e.g.
Willaman & Schubert, 1961), regarding the occurrence
of caeine may give the impression that this purine al-
kaloid is shared by a large number of genera.
However, if we rely exclusively on data re-examined by
advanced analytical techniques, we arrive at the con-
clusion that during evolution, 'invention` of caeine,
i.e. the purine alkaloid pathway, was a relatively rare
event meaning that out of ca. 10,000 angiosperm gen-
era only seven developed, to our present knowledge,
this phytochemical feature, namely Coea, Camellia,
Theobroma, Herrania, Cola, Ilex, and Paullinia.
Therefore, a report by Stewart in 1985 (Stewart, 1985)
was most exciting because it claimed the presence of
caeine, even though in the very low range of 6 and
50 ppm (31 and 258 nmol g
ÿ1
fr. wt) in leaves and
¯owers, respectively, of several Citrus species. His ®nd-
ings were con®rmed by an Italian group (Trova,
Cossa, & Gandolfo, 1994) which detected caeine in
dried citrus ¯owers (237 to 856 nmol g
ÿ1
dry wt) com-
mercially available for preparing a tea. Moreover, the
authors found caeine for the ®rst time also in honeys
originating from the activities of honey-bees (Apis mel-
lifera ) visiting either frequently (uni¯oral honey) or
sporadically ('mille®ori` honey) the ¯owers in orange
plantations. Caeine in these honeys ranged from 2.6
to 52 nmol g
ÿ1
. Later, the analyses were extended to
various other citrus honeys (De®lippi, Piancone, &
Tibaldi, 1995; Vacca & Fenu, 1996) and possible
source-¯owers (Vacca, Agabbio, & Fenu, 1997) with
the aim to establish a measure of quality control.
However, a correlation could not yet be established. In
all these studies, neither single ¯ower organs nor nec-
tar and pollen were examined.
In a preliminary investigation on coee (Kretschmar
& Baumann, 1998) we recognised that ¯ower caeine
was relatively abundant (ca. 3200 nmol g
ÿ1
fr. wt) and
preferentially, even though not very markedly, allo-
cated in the androecium together with other purine al-
kaloids. Therefore, we analysed the within-the-¯ower
distribution of these alkaloids in citruses (including
Poncirus ) considered ideal to demonstrate organ-
speci®c allocation because of the low average content.
Indeed, among the ¯ower organs analysed the androe-
cium had by far the highest concentration of alkaloids,
Phytochemistry 52 (1999) 19±23
0031-9422/99/$ - see front matter #1999 Elsevier Science Ltd. All rights reserved.
PII: S0 0 3 1 - 9 4 2 2 ( 9 9 ) 0 0 1 1 9 - 3
* Corresponding author. Fax: +41-1-634-8204.
E-mail address: twb@botinst.unizh.ch (T.W. Baumann)

with a main allocation in the anther and pollen at a
very high level. Moderate, but still marked purine al-
kaloid concentrations were found in the nectar. Since
caeine is known to be insecticidal, the results were
discussed also in the context of intoxication of honey-
bees.
2. Results and discussion
In a ®rst approach the purine alkaloid content of
entire ¯owers of Citrus paradisi and C. maxima was
determined. Caeine was the main alkaloid (21 and 77
nmol g
ÿ1
fr. wt, respectively) accompanied by theo-
phylline (4 and 15 nmol g
ÿ1
fr. wt, respectively), while
theobromine and paraxanthine occurred in traces only.
Similarly, commercially available orange ¯ower tea (C.
sinensis ) contained caeine (182 nmol g
ÿ1
dry wt) and
theophylline (46 nmol g
ÿ1
dry. wt). These caeine
values are in the range as found before (Stewart, 1985;
Trova et al., 1994), but until now the presence of theo-
phylline and other dimethylxanthines in Citrus has not
been reported. Theophylline is a trace compound in
the `classical' caeine plants consumed by the human.
In some of the citrus ¯ower tissues it showed consider-
able accumulation (Table 1).
The chemical analysis of ¯ower development (C.
limon ) revealed, that the small, round-shaped ¯ower
bud was virtually alkaloid-free, with only a trace of
theophylline, whereas the elongated bud shortly before
anthesis contained well-measurable concentrations of
both caeine and theophylline, accompanied by little
theobromine and paraxanthine. During anthesis the
caeine content increased by a factor of almost 2
(Table 1). Similarly (not shown), two stages (145 and
216 mg fr. wt) of ¯ower buds of the closely related
Poncirus trifoliata were purine alkaloid-free, whereas
the freshly opened ¯ower (375 mg fr. wt) had an over-
all caeine concentration of 109 nmol g
ÿ1
fr. wt. A
similar increase was observed during anthesis of C.
paradisi. Therefore we may conclude that anthesis in
citruses is coordinated with a rapid allocation of caf-
feine.
Then, the ¯owers of C. paradisi,C. maxima,,C.
limon and P. trifoliata were separated into petals, pis-
tils and stamens and analysed (the tiny green sepals
and the ¯ower base were found in preliminary exper-
iments to be virtually alkaloid-free). Both, petals and
pistils contained very small concentrations (mostly in
the range of 2 to 10 nmol g
ÿ1
fr. wt, respectively) of
caeine, theobromine, or theophylline (not shown).
However, the stamens (Table 1) contained the highest
concentrations of caeine and theophylline and were
the exclusive site of ¯ower paraxanthine which was
hardly detectable in the related entire ¯owers because
of dilution. The separate analysis of ®lament and
anther revealed maximum alkaloid concentrations in
the latter exceeding altogether the concentration of
Table 1
Purine alkaloid content (nmol g
ÿ1
fr. wt) in ¯owers, nectars and honeys of Citrus spp. n= 3±10 (¯owering units); n.d.=not detectable
Caeine Theobromine Theophylline Paraxanthine
Flower development (C. limon)
Small bud (148 mg; n= 5) n.d. n.d. 621
a
n.d.
Large bud (840 mg; n= 3) 16622
a
252251
Full anthesis (938 mg; n= 3) 318231 27211
Stamens
C. paradisi 3'233242 28 1732222
C. maxima 1'110213 13 305 245
P. trifoliata 807211 5 56215
C. limon 1'415218 67 115228
Filament
C. paradisi 1'917229 18 1392215
C. maxima 850211 8 12 214
Anther
C. paradisi 8'5512169 26 1'491229 32
C. maxima 7'9002119 18 2'753 241 20
Pollen
C. medica 6'8572321 n.d. 1'921290 n.d.
Nectar (nmol ml
ÿ1
)
C. paradisi 487215 22 552212
C. maxima 912510 3213
C. limon 6022 n.d. n.d. n.d.
Honey (nmol g
ÿ1
)
Sicilia 31216 3213
California 221 1 n.d. n.d.
a
Mean value (2the estimated experimental error, see Section 3).
J.A. Kretschmar, T.W. Baumann / Phytochemistry 52 (1999) 19±2320

10,000 nmol (=10 mmol) per g fr. wt. If related to dry
wt, it results a value of ca. 0.7±0.8% caeine and
0.9% total purine alkaloid in the anther. Hence, the
purine alkaloid concentration in the anther is close to
that in the Arabica coee bean (1.2%, almost exclu-
sively caeine)!
Finally, pollen of (due to the absence of blooming
of the other species) Citrus medica (citron) was ana-
lysed. This single analysis revealed a high purine alka-
loid concentration (altogether almost 8800 nmol g
ÿ1)
in the microspores (Table 1), which is in the range of
that found in the anther of the closely related species.
We cannot yet decide whether the anther alkaloid
amount is completely con®ned to the pollen, or
whether the anther wall contains alkaloid at concen-
trations similar to pollen. Also, we have not yet stu-
died the localisation of purine alkaloids within the
pollen grain. At the present we can only speculate
about the signi®cance of this conspicuous allocation.
Besides protection against (unknown) pollen predators,
purine alkaloids are well-studied defence compounds
(reviewed in Harborne (1993)), the cytokinin-like eect
of caeine (Vito
Âria & Mazzafera, 1997) may play a
role during pollination and seed set in citrus
(Hernandez Minana & Primo Millo, 1990).
Blossom honey essentially consists of nectar concen-
trated by the activities of specialised bees in the hive.
The pollen present in the honey is quantitatively negli-
gible but a valuable nectar 'contaminant` which is of
help to trace the source ¯owers. It has been reported
that on average 64% of the pollen found in uni¯oral
citrus honeys is citrus pollen (White & Bryant, 1996)
meaning that roughly two third of the honey originate
from citrus nectar. Since the latter was shown to con-
tain caeine in the range of ca. 60 to 490 nmol ml
ÿ1
(Table 1) and undergoes a concentration process by a
factor of ca. 2 during honey production, one should
expect a much higher caeine concentration in citrus
honey than found in the present (2 and 31 nmol g
ÿ1
;
Table 1) or in earlier (2.5±50 nmol g
ÿ1
(Trova et al.,
1994)) studies. It can be calculated that about 95% of
the nectar caeine is removed or degraded by a still
unknown mechanism. However, even though the nec-
tar was sampled with caution we cannot rule out con-
tamination by pollen.
In order to obtain an estimate of chemical defence
allocation, the number and weight of the individual
¯ower organs were determined and the alkaloid distri-
bution calculated as exempli®ed for C. lemon in Table
2. In summary, one citrus ¯ower contains about 260
nmol (ca. 50 mg) caeine and 300 nmol total purine al-
kaloid, 99 and 96.6%, respectively, allocated to the
androecium! The amount of alkaloids found in the
nectar (ca. 20 ml) at the moment of ¯ower dissection is
negligible in the case of C. lemon (ca. 1 nmol; caeine
only), but was distinctly higher in C. paradisi (ca. 10
nmol) and may be considerable if extrapolated to the
entire period of ¯owering. However, nectar secretion
was not studied in detail. Diurnal ¯uctuations are to
be expected and may account for some of the dier-
ences in pattern and concentrations of nectar purine
alkaloids listed in Table 1.
Finally, we should mention that orange ¯ower tea
(Aurantii ¯os) is pharmaceutically recommended for
treatment of sleeplessness [O
ÈAB, Ph. Helv. VI]. The
amount of caeine ingested by consumption of such a
calming tea is below 100 mg, a dose present in homeo-
pathic coee preparations used against insomnia
(HAB, (Baumann & Seitz, 1992)).
2.1. Caeine and honey-bees
High concentrations of secondary compounds have
been detected in microspores of both wind- (e.g.
Meurer, Wray, Wiermann, & Strack, 1988) and insect-
(reviewed in Detzel & Wink, 1993) pollinated ¯owers.
In the latter case the question of intoxication of the
pollinators arises. Detzel and Wink (Detzel & Wink,
1993) tested a large number of such allelochemicals on
the feeding behaviour of honey-bees. Caeine was
found to act as a deterrent and its toxicity was com-
paratively low under no-choice conditions (LD
50
at 0.2
%). In an earlier study (Ishay & Paniry, 1979), honey-
bees, oered free choice of either the sugar solution
alone or the sugar solution with the caeine, similarly
preferred the sugar solution. The concentration of the
Table 2
Purine alkaloid allocation within the ¯ower of Citrus lemon. The ¯ower base is virtually alkaloid-free
Organ Caeine aPurine alkaloids
fr. wt (mg) nmol g
ÿ1
fr. wt amount (nmol) % of total nmolg
ÿ1
fr. wt Amount (nmol) % of total
Petals (4±5) 509 3.9 2.0 0.8 15.3 7.8 2.5
Stamens (31±32) 183 1415.0 258.9 99.1 1604.5 293.6 96.5
Pistil (1) 103 2.5 0.3 0.1 29.5 3.0 1.0
Flower base (1) 119 n.d. 0 ± ± ± ±
Total 914 285.8
a
261.2 100 333.0
a
304.4 100
a
Overall concentration of the entire ¯ower. In parenthesis: number of organs per one ¯ower. For nectar see text.
J.A. Kretschmar, T.W. Baumann / Phytochemistry 52 (1999) 19±23 21

caeine solution (ca. 250 mM) happened to be in the
range of citrus nectars (Table 1). It was readily
accepted under no-choice conditions. After ®ve days, a
300±500 % boost in oviposition by the (young) queen,
an enhanced activity of the bees outside the hive, and
an improved defence by bees against hornets at the
hive entrance was observed. In contrast, hornets also
fed with caeine ceased foraging in the ®eld and also
failed to clear the dead (poisoned by caeine) larvae
out of the nest. On the cellular level, caeine was
shown to in¯uence the cytosolic Ca
2+
concentration in
various organisms including honey-bees, where the caf-
feine-sensitive Ca
2+
release from the endoplasmic reti-
culum in photoreceptors has been studied (Walz,
Baumann, & Ciriacy-Wantrup, 1994).
Caeine and related substances are known not only
to exert insecticidal activity but also to synergize the
eects of pesticides (Nathanson, 1984). On the one
side it appears that the toxicity of caeine in the
honey-bee larvae, which to our knowledge do not suf-
fer from the caeine-rich pollen in citrus orchards, is
relatively low. On the other side, however, one should
in future evaluate bee toxicity of pesticides also in the
presence of purine alkaloids, in order to account for
the above-mentioned synergistic eect which may
occur not only in citrus but also in coee and tea plan-
tations.
Finally, we should mention that recently citrus pol-
len was found to be toxic to the predatory mite
Euseius mesembrinus (Yue, Childers, & Fouly, 1994)
which is a facultative pollen feeder widely distributed
in citrus plantations. If toxicity is due to the presence
of purine alkaloids, one may suggest that mites, in
contrast to honey-bees, are very susceptible to purine
alkaloids, a situation which could be advantageous in
the chemical control of the ectoparasitic mite Varroa
jacobsoni associated with the honey-bee.
3. Experimental
3.1. Plant material
Citrus plants (C. paradisi Macf., grapefruit; C. max-
ima (Burm.) Merr., shaddock; C. limon (L.) Burm.f.,
lemon, and C. medica L., citron) were grown in the
greenhouse. The trifoliate orange plant (P. trifoliata
Raf.) was kept outdoors in the institute garden. Tea of
orange ¯owers (C. sinensis (L.) Pers. was bought in a
local store (Coop, Switzerland) as well as orange
¯ower honeys of Sicilian (Globus, Switzerland) and
Californian (Biorex AG, Ebnat-Kappel, Switzerland)
origin. Nectar was sampled using a glass (hematokrit)
capillary.
3.2. Purine alkaloid extraction
Fresh entire ¯owers (n= 3 to 10) or the related
¯ower parts were pooled and extracted in 0.1 N HCl
(1 ml per 50 to 150 mg fr. wt) at 508for 30 min by
sonication. One ml of the extract was applied onto a
Kieselgur column (Extrelut1, Merck). Essential oil was
removed by 12 ml hexane and thereafter purine alka-
loids were eluted with 12 ml CH
2
Cl
2
. The eluate was
dried under a stream of N
2
and the residue dissolved,
for HPLC, in 1 ml H
2
O. Orange ¯ower tea was
extracted likewise (1 ml 0.1 N HCl per 250 mg, with-
out further drying the ¯owers) as well as honey (1 g),
which was diluted with 0.1 N HCl (2 ml) prior to ap-
plication onto the Kieselgur column. Nectar was
mixed (1:1) with 8% MeOH and directly injected.
Pollen was collected and processed as follows: stamens
were harvested, dried at room temp. and transferred
into a pre-weighed Eppendorf tube, which then was
vortexed at high speed to spin o the pollen. The sta-
mens were removed and the weight of the pollen was
determined by weighing the tube again (Mettler AE
240). The pollen (2.84 mg) was suspended in 300 ml 0.1
N HCl. After 3 h at room temp. a 30-min-sonication
at 408followed. Thereafter, the suspension was ®ltered
through a membrane ®lter and directly injected into
HPLC. No attempt was made to determine the dry
weight of the pollen.
3.3. HPLC separation of purine alkaloids
HPLC separation of purine alkaloids was carried
out on a Nucleosil-100-5 C18 HD column; precolumn
48 mm; ChromCart, Macherey-Nagel). Parameters
were controlled by a Hewlett-Packard liquid chromato-
graph equipped with a diode array detector set at 272
nm. Chromatography was carried out using the follow-
ing gradient: 0±4 min with 0±7.5% MeOH and 0±
2.5% AcN, 4±20 min with 7.5% MeOH and 2.5%
AcN. The R
t
's (min) were 7.1, 9.1, 9.6 and 16.0 for
theobromine, paraxanthine, theophylline, and caeine,
respectively. The ¯ow rate was 1.1 ml min
1
and injec-
tion vol 150 ml. Peak identi®cation was achieved by
comparing UV spectrum (library established under
separating conditions) and retention time of authentic
standards.
3.4. Calculation of the mean value and the `systematic
experimental error'
Since in this study we did not aim at showing the
variability of purine alkaloids and their concentrations
among ¯owers from one or several plants, the ¯owers
were randomly collected and pooled (with ngener-
ally=10, in a few cases of shortage lower but never
less than 3). The obtained mean value has an exper-
J.A. Kretschmar, T.W. Baumann / Phytochemistry 52 (1999) 19±2322

imental error which was calculated by considering the
accuracy of the fresh weight determination (21 mg)
and the quantitation by HPLC (21%).
Acknowledgements
We thank the beekeepers Kurt Honegger, Baden
and Pierre Pittier, Dietlikon, for answering awkward
questions of botanists and Stefan Bogdanov, Federal
Dairy Research Institute, Bee Department, Bern
Liebefeld, Switzerland, for his expertise in bee-related
topics as well as for critically reading the manuscript.
This work was ®nancially assisted by the Swiss
National Science Foundation, Grant No. 31-50521.97
and by the `Jubila
Èumsspende fu
Èr die Universita
Èt
Zu
Èrich'.
References
Baumann, T. W., & Seitz, R. (1992). In R. Ha
Ènsel, K. Keller, H.
Rimpler, & G. Schneider, (p. 926). In Hagers Handbuch der
Pharmazeutischen Praxis,Vol. 4. Berlin: Springer-Verlag.
De®lippi, A., Piancone, G., & Tibaldi, G. (1995). Industrie
Alimentari,34,6.
Detzel, A., & Wink, M. (1993). Chemoecology,4,8.
Harborne, J. B. (1993). Ecological biochemistry. London: Academic
Press.
Hernandez Minana, F. M., & Primo Millo, E. (1990). Journal of
Horticultural Science,65, 595.
Ishay, J. S., & Paniry, V. A. (1979). Psychopharmacology,65, 299.
Kretschmar, J. A., & Baumann, T. W. (1998). Future trends in phyto-
chemistry, PSE ± symposium. The Netherlands: Rolduc.
Meurer, B., Wray, V., Wiermann, R., & Strack, D. (1988).
Phytochemistry,27, 839.
Nathanson, J. A. (1984). Science,226, 184.
Stewart, I. (1985). Journal of Agricultural and Food Chemistry,33,
1163.
Trova, C., Cossa, G., & Gandolfo, G. (1994). Industrie Alimentari,
33, 403.
Vacca, V., & Fenu, P. (1996). Industrie Alimentari,35, 368.
Vacca, V., Agabbio, M., & Fenu, P. (1997). Industrie Alimentari,36,
611.
Vito
Âria, A. P., & Mazzafera, P. (1997). Biologia Plantarum,40, 329.
Walz, B., Baumann, O., & Ciriacy-Wantrup, E. V. (1995). Journal of
General Physiology,105, 537.
Willaman, J. J., & Schubert, B. G. (1961). Agric. Res. Serv. US
Dept. Agric. Techn. Bull.,1234,1.
White, J. W., & Bryant, V. M. J. (1996). Journal of Agricultural and
Food Chemistry,44, 3423.
Yue, B., Childers, C. C., & Fouly, A. H. (1994). International
Journal of Acarology,20, 103.
J.A. Kretschmar, T.W. Baumann / Phytochemistry 52 (1999) 19±23 23